Advancing Artificial Intelligence for Remote Sensing: Overcoming Data Scarcity and Domain Shift

Ma, Yuchi

IN005
Advancing Artificial Intelligence for Remote Sensing: Overcoming Data Scarcity and Domain Shift

Session ID#: 251125

Session Description:

Artificial Intelligence (AI) has proven powerful for utilizing the increasing amounts of remote sensing (RS) data for applications such as landcover mapping, crop yield prediction, and water resources management. However, many AI models require a substantial amount of ground truth labels for training. Collecting these labels often involves pixel-wise image labeling or in-field sample collection, which can demand significant labor and financial resources. Due to the distribution shift and spatial heterogeneity among regions, models trained in one region often demonstrate poor performance when applied to other regions. Therefore, it is important to develop approaches to address labeled data scarcity and effectively train AI models for RS applications. This session welcomes studies and applications of transfer learning, self-supervised learning, and foundation models on RS that improve performance under domain shift or limited data, domain knowledge-guided methods to reduce the need for real-world data samples, and benchmark datasets in RS.

Co-Sponsor(s):

B - Biogeosciences
EP - Earth and Planetary Surface Processes
SY - Science and Society

Index Terms:

0480 Remote sensing [BIOGEOSCIENCES]
1906 Computational models, algorithms [INFORMATICS]
1926 Geospatial [INFORMATICS]
1942 Machine learning [INFORMATICS]

Primary Convener: Yuchi Ma, Stanford University, Department of Earth System Science & Center on Food Security and the Environment, Stanford, CA, United States

Conveners: Sherrie Wang, Massachusetts Institute of Technology, Mechanical Engineering, Cambridge, United States and Hannah Rae Kerner, Arizona State University, School of Computing and Augmented Intelligence, Tempe, AZ, United States

Student/Early Career Convener: Siddharth Sachdeva, Organization Not Listed, Washington, United States; Stanford University, Stanford, United States

Deep learning-based gap-filling and agricultural field boundary delineation empowers early-season rice paddy mapping in smallholder farming system (1847904)

Ryoungseob Kwon^1,2, Youngryel Ryu³, Changhyun Choi⁴, Pengfei Tang⁴, Seongwoo Jo³ and Kyungdo Lee⁵, (1)Seoul National University, Integrated Major in Smart City Global Convergence, Seoul, Korea, Republic of (South), (2)Seoul National University, Interdisciplinary Program in Landscape Architecture, Seoul, Korea, Republic of (South), (3)Seoul National University, Department of Landscape Architecture and Rural Systems Engineering, Seoul, Korea, Republic of (South), (4)Seoul National University, Research Institute of Agriculture and Life Sciences, Seoul, South Korea, (5)National Agricultural Satellite Center, National Institute of Agricultural Sciences, Jeonju-si, South Korea

A GAN-Enhanced Deep Learning Framework for Rooftop Detection from Historical Aerial Imagery (1852689)

Pengyu Chen¹, Sicheng Wang² and Cuizhen Wang², (1)University of South Carolina Columbia, Columbia, SC, United States, (2)University of South Carolina, Columbia, United States

Mapping Mining Footprints with Spatiotemporal CNNs and Multi-Head Self-Attention Mechanism: A Satellite Remote Sensing and Deep Learning Approach. (1859424)

Ikramul Hasan and Desheng Liu, The Ohio State University, Department of Geography, Columbus, United States

Multi-Modal Remote Sensing Segmentation of Geomorphological Structures with Deep Learning and Self-Supervision (1863801)

François Faure¹, Lorena Rosell-Guevara², Cerine Bencheikh El Amira¹, Swann Zerathe¹, Audin Laurence¹, James Hollingsworth¹ and Sophie Giffard-Roisin¹, (1)Univ. Grenoble Alpes, Univ. Savoie Mont Blanc, CNRS, IRD, Univ. Gustave Eiffel, ISTerre, Grenoble, France, (2)Instituto Geológico, Minero y Metalúrgico, INGEMMET, San Borja, Peru

Lightweight Semantic Segmentation for Co-seismic Landslide Identification Using Transfer Learning (1864758)

Shaoqiang Meng^1,2, Zhenming Shi¹ and Thomas Glade², (1)Tongji University, Shanghai, China, (2)University of Vienna, Vienna, Austria

Corn Yield Prediction Using Phenology-Informed Remote Sensing Data and Deep Learning Methods (1868809)

Mingda Wu, Xiaoyu Wang and Zhou Zhang, University of Wisconsin-Madison, Biological Systems Engineering, Madison, United States

Towards Multi-Sensor and Multi-Task Generative Foundation Model for Earth Observation (1872132)

Mr. Jiancheng PAN and Xiaomeng Huang, Tsinghua University, Department of Earth System Science, Beijing, China

A Deep Learning Framework for Tree Detection in Forest Point Clouds Using Multi-Layered Forest Structure (1877727)

Yiliu Tan¹, Xin Yang¹, Jingyi Zhang², Xin Xu³ and Maiko Shigeno⁴, (1)University of Tsukuba, Graduate School of Systems and Information Engineering, Tsukuba, Japan, (2)University of Tsukuba, Center of Research in Radiation, Isotopes, and Earth System Sciences, Tsukuba, Japan, (3)University of Maryland at College Park, Department of Geographical Sciences, Maryland, United States, (4)University of Tsukuba, Institute of Engineering, Information and Systems, Tsukuba, Japan

Urban-Trained, Forest-Ready: A Source-Mix Domain-Adaptation Pipeline for Large-Scale Forest Point Cloud Segmentation (1878597)

Fei Zhang¹, Robert Ormal Chancia², Amirhossein Hassanzadeh³ and Jan van Aardt³, (1)Rochester Institute of Technology, Chester F. Carlson Center for Imaging Science, Rochester, United States, (2)Rochester Institute of Technology, Imaging Science, Rochester, NY, United States, (3)Rochester Institute of Technology, Imaging Science, Rochester, United States

Mapping Crop Field Size Change in Conflict Hotspots of Sub-Saharan Africa Using Very-High-Resolution Images and Limited Reference Data (1881689)

Aleksandra Wasik¹, Philippe Rufin^2,3, Esther Ibrahim Dr^4,5, Lyndon D Estes⁶, Tanmoy Chakraborty⁷, Fabien Cottier Dr⁸ and He Yin¹, (1)Kent State University, Department of Geography, Kent, United States, (2)Universite catholique de Louvain (UCLouvain), Earth and Life Institute, Louvain-la-Neuve, Belgium, (3)Humboldt University of Berlin, Geography Department, Berlin, Germany, (4)Leibniz Centre for Agricultural Landscape Research, Müncheberg, Germany, (5)National Space Research and Development Agency, Abuja, Nigeria, (6)Clark University, Graduate School of Geography, Worcester, MA, United States, (7)Clark University, Graduate School of Geography, Worcester, United States, (8)Columbia University, Center for Integrated Earth System Information (CIESIN), Climate School, Palisades, United States

Scalable Early-Season Soybean Mapping Without Ground Truth Labels: Unsupervised Integration of Physiological Traits and Spectral-Temporal Dynamics (1885246)

Hyeyoung Jo¹, Youngryel Ryu¹, Ryoungseob Kwon², Pengfei Tang³, Weiguo Yu³ and Dr. Changhyun Choi⁴, (1)Seoul National University, Department of Landscape Architecture and Rural Systems Engineering, Seoul, South Korea, (2)Seoul National University, Interdisciplinary Program in Landscape Architecture, Seoul, Korea, Republic of (South), (3)Seoul National University, Research Institute of Agriculture and Life Sciences, Seoul, South Korea, (4)Seoul National University, Research Institute of Agriculture and Life Sciences, Seoul, Korea, Republic of (South)

Developing a Neighborhood-Relation Poverty Mapping Method Using Household Survey and Satellite Imagery (1888890)

Tunyu Chang, University of Tokyo, Bunkyo-ku, Japan and Akiyuki Kawasaki, The University of Tokyo, Institute for Future Initiatives; Department of Civil Engineering; Institute for Digital Observatory., Tokyo, Japan; The University of Tokyo, Department of Civil Engineering, Tokyo, Japan; The University of Tokyo, Institute for Digital Observatory, Tokyo, Japan

SpectraScope: Spectral‑Conditioned Foundation Models for Label‑Efficient, Domain‑Robust Remote‑Sensing Analytics (1904822)

Junghwan Park, TelePIX Inc., AI Research Team, Seoul, South Korea and Darongsae Kwon, TelePIX Inc., Seoul, South Korea

Improving Small Vessel Detection in Satellite Imagery via Pretraining and Architecture Refinement under Multispectral Data Scarcity (1907092)

Seon-Hoon Kim¹, Yerin KIM², Hyeji Sim², Mi-Hyeon Jeong¹ and Daewon Chung², (1)University of Science and Technology (UST), Daejeon, South Korea, (2)Korea Aerospace Research Institute (KARI), Daejeon, South Korea

Global Agricultural Field Boundaries Based on the Segment Anything Model (1915139)

Juliana de Abreu Araujo, Mississippi State University, Mississippi State, MS, United States, Vitor Martins, Mississippi State University, Department of Agricultural and Biological Engineering, Mississippi State, MS, United States, Lucas Borges Ferreira, Mississippi State University, North Mississippi Research and Extension Center, Tupelo, MS, United States and Uilson R. V. Aires, Mississippi State University, Starkville, United States

Learning with less: label-efficient land cover classification at very high spatial resolution using semantic stratification and self-supervised deep learning (1916296)

Dakota Hester¹, Vitor Martins¹, Lucas Borges Ferreira² and Thainara Lima¹, (1)Mississippi State University, Department of Agricultural and Biological Engineering, Mississippi State, MS, United States, (2)Mississippi State University, North Mississippi Research and Extension Center, Tupelo, MS, United States

A Physics-Informed, AI-Driven System for Traffic Estimation (1918584)

Tewodros Gebre, North Carolina A & T State University, Greensboro, NC, United States and Leila Hashemi Beni, North Carolina A & T State University, Department of Built Environment, Greensboro, United States

On the Foundations of Earth and Climate Foundation Models (Invited) (1926202)

Xiao Xiang Zhu, Technical University of Munich, Munich, Germany

A Reproducible Open-Source Machine Learning Framework for Sentinel‑2 Land Use and Land Cover Classification: Comparative Analysis of Random Forest and Support Vector Machine Methods (1933044)

Md. Jubayer Ahamed, Pranto Kumar Sarker and Sujoy Dey, Bangladesh University of Engineering and Technology, Department of Water Resources Engineering, Dhaka, Bangladesh

Parameter-Efficient Fine-Tuning of Vision-Language Models for High-Performance Earth Observation Tasks (1936359)

Weijia Wang¹, Jichao Fang¹ and Wei Luo², (1)Northern Illinois University, DeKalb, IL, United States, (2)Northern Illinois Univeristy, DeKalb, United States

WildSAT: Learning Satellite Image Representations from Wildlife Observations (1936433)

Rangel Daroya¹, Elijah Cole², Oisin Mac Aodha³, Grant Van Horn⁴ and Subhransu Maji⁴, (1)University of Massachusetts Amherst, College of Information and Computer Sciences, Amherst, MA, United States, (2)GenBio AI, Palo Alto, United States, (3)University of Edinburgh, School of Informatics, Edinburgh, United Kingdom, (4)University of Massachusetts Amherst, College of Information and Computer Sciences, Amherst, United States

Assessing Performance of Remote-Sensing Based Machine Learning Approaches for Soil Moisture Retrieval Across Agricultural Landscapes Using ISMN (1936692)

Veda Sunkara, Stanford University, Stanford, CA, United States, Siddharth Sachdeva, Stanford University, Stanford, United States, David B Lobell, Stanford University, Department of Earth System Science & Center on Food Security and the Environment, Stanford, United States and Alexandra G Konings, Stanford University, Earth System Science, Stanford, United States

Multi-Polarization SAR Super-Resolution with a Spatially Aware Hybrid GAN-U-Net Architecture (1947957)

Jagrati Talreja, North Carolina A & T State University, Greensboro, NC, United States and Leila Hashemi Beni, North Carolina A & T State University, Department of Built Environment, Greensboro, United States

Characterization of Strike-Slip Fault Offsets Using Deep Learning and Simulated Topographic Data (1948213)

Jules Bourcier¹, Sarah Visage², Sofia Goncalves¹, Margaux Mouchené¹, Léa Pousse Beltran¹ and Sophie Giffard-Roisin¹, (1)Institut des Sciences de la Terre, Université Grenoble Alpes, Grenoble, France, (2)Sapienza University of Rome, Istituto di Geologia Ambientale e Geoingegneria - CNR c/o Dipartimento di Scienze della Terra, Rome, Italy

Combining open labeled datasets with varying domains to improve large-scale agricultural field boundary delineation (1952671)

Rufai Omowunmi Balogun¹, Tanmoy Chakraborty², Gedeon Muhawenayo³, Hannah Rae Kerner³, Ana M Tarano³, Zhanpei Fang⁴, Nathan Jacobs⁵, Subash Khanal⁵, Rahebe Abedi⁶ and Lyndon D Estes², (1)Clark University, Clark Center for Geospatial Analytics, Worcester, MA, United States, (2)Clark University, Graduate School of Geography, Worcester, United States, (3)Arizona State University, School of Computing and Augmented Intelligence, Tempe, AZ, United States, (4)Oregon State University, Geography Program, College of Earth, Ocean, and Atmospheric Sciences, Corvallis, United States, (5)Washington University in St. Louis, Department of Computer Science and Engineering, St Louis, MO, United States, (6)Clark University, Graduate School of Geography, Worcester, MA, United States

Deep Learning-Based Prediction of Soil Organic Carbon Using Remote Sensing Data in an Intensively Farmed Agroecosystem: A Hindon Basin Case Study (1959244)

Sai Bargav Reddy, Indian Institute of Technology Roorkee, Roorkee, India and Vinnarasi R, Indian Institute of Technology Roorkee, Civil Engineering Department, Roorkee, India

Mapping Fallow Fields from Sparse and Fuzzy Positive-Only Labels: Implications for Food-Water Nexus (1959597)

Sk Muhammad Asif and Orhun Aydin, Saint Louis University, Earth, Environmental and Geospatial Science, Saint Louis, United States

Toward Reliable Global Monitoring of Vegetation Cover Using Spatial-Temporal Adjustment of Machine Learning Models (1963985)

Bo Zhou, University of California Los Angeles, Los Angeles, CA, United States and Greg S Okin, University of California Los Angeles, Department of Geography, Los Angeles, United States

An Interface for Data Curation and Mapping of Irrigated Areas Using Active Learning (1964082)

Kin NG¹, Jack Marquez¹, Gabriel Laboy¹, Endalkachew Abebe Kebede², Holly A Michael³, Kyle Frankel Davis⁴ and Michela Taufer¹, (1)University of Tennessee, Department of Electrical Engineering & Computer Science, Knoxville, United States, (2)University of Delaware, Department of Geography & Spatial Sciences, Newark, DE, United States, (3)University of Delaware, Earth Sciences, Civil and Environmental Engineering, Newark, DE, United States, (4)University of Delaware, Geography & Spatial Sciences, Newark, DE, United States

Bayesian Deep Learning for High-Resolution Bathymetry Retrieval in Data-Scarce Harbors and Uncertainty-Driven Coastal Wave Modeling (1971356)

Nadya Chakir¹, Abdellah El Kacimi¹, Mohammed Seaid² and Nabil El Mocayd³, (1)FP Safi, Cadi Ayyad University, Laboratory of Modeling and Combinatorial, Safi, Morocco, (2)University of Durham, Department of Engineering, Durham, United Kingdom, (3)Mohammed VI Polytechnic University (UM6P), International Water Research Institute (IWRI), College of Agriculture & Environmental Sciences (CAES), Ben Guerir, Morocco

Bridging Spatial Gaps in Crop Yield Mapping to Support Climate-Resilient Agriculture via Knowledge-Guided Graph Neural Networks (1971569)

Jie Yang¹, Licheng LIU², Qi Yang³, Xiaowei Jia⁴, Bin Peng¹, Kaiyu Guan⁵ and Zhenong Jin², (1)University of Illinois Urbana-Champaign, Urbana, United States, (2)University of Minnesota Twin Cities, Department of Bioproducts and Biosystems Engineering, Saint Paul, United States, (3)University of Minnesota Twin Cities, Department of Biogeochemical Integration, Max Planck Institute for Biogeochemistry, Jena, Germany, (4)University of Pittsburgh, Pittsburgh, United States, (5)University of Illinois Urbana-Champaign, Agroecosystem Sustainability Center, Institute for Sustainability, Energy, and Environment, Urbana, United States

Peruvian Land Cover Characterization using Machine Learning and Remote Sensing (1974419)

Daria-Ioana Radu¹, Hugo Lepage¹, Eustace Barnes¹, Milagros Villar², Michael Huisa², John Forrest¹ and Crispin H.W. Barnes¹, (1)University of Cambridge, Cambridge, United Kingdom, (2)Universidad Nacional Autonoma de Huanta, Environmental Science, Huanta, Peru

Spatially Adaptive GeoAI Framework for Multi-Scale Environmental Exposure Assessment by using Satellite and Street View Imagery (1976746)

Abdulla Al Kafy, University of Texas at Austin, Geography and Environment, Austin, United States and Kelley Crews, University of Texas at Austin, Geography and Environmnet, Austin, United States

Empowering Wildfire Intelligence: Automated Segmentation, Instance Tracking, and Benchmarking with SIT-FUSE and Multi-Sensor Observations (1977573)

Nicholas LaHaye¹, Lily McKenna², Anastasija Easley³, Andrei Zagvozdkin⁴, Igor Yanovsky⁵, Ryan DeMilt⁶, Myscon Truong¹, Noah Giebink⁶, Hugo Lee⁵, Kyongsik Yun⁵, Yuzhou Chen⁷, Yulia Gel⁴, Olga V. Kalashnikova⁵ and David S Saah⁸, (1)Spatial Informatics Group, LLC, Alameda, United States, (2)University of Arizona, Wyant College of Optical Sciences, Tucson, United States, (3)University of California Berkeley, Berkeley, CA, United States, (4)Virginia Polytechnic Institute and State University, Department of Statistics, Blacksburg, United States, (5)NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, United States, (6)Spatial Informatics Group, LLC, Alameda, CA, United States, (7)University of California Riverside, Department of Statistics, Riverside, United States, (8)University of San Francisco, Geospatial Analysis Lab, San Francisco, CA, United States

Remote Sensing-Based Multi-Temporal Landslide Inventory Mapping: Comparing GIS Thresholding and Random Forest Classification on LiDAR DEMs in Baltimore County, Maryland. (1981374)

Raymond Olamijulo¹, Micheal Okegbola², Oludare Owolabi² and Samuel Fakolade¹, (1)Morgan State University, Baltimore, MD, United States, (2)Morgan State University, Civil and Environmental Engineering, Baltimore, United States

Enhancing Climate Literacy in Engineering, Integrating Climate Resilience into Education and Practice. (1983087)

Samuel Fakolade¹, Oludare Owolabi², Micheal Okegbola² and Raymond Olamijulo¹, (1)Morgan State University, Baltimore, MD, United States, (2)Morgan State University, Civil and Environmental Engineering, Baltimore, United States

Benchmarking Vision Foundation Models for Automated Landslide Inventory Mapping (1985487)

Weiwei Zhan, University of Central Florida, Department of Civil Environmental and Construction Engineering, Orlando, FL, United States and Chen Chen, University of Central Florida, Center for Research in Computer Vision, Orlando, United States

Modeling drought-based extreme events using a grounded N-shot learning approach (1988735)

Shrey Gupta, Organization Not Listed, Washington, DC, United States, Yi Ming, Boston College, Chestnut Hill, United States and George Mohler, Boston College, MA, United States

Invariant Features for Global Crop Type Classification (2002041)

Xinyi Tong¹ and Sherrie Wang^1,2, (1)Massachusetts Institute of Technology, Laboratory for Information and Decision Systems, Cambridge, United States, (2)Massachusetts Institute of Technology, Mechanical Engineering, Cambridge, United States

See more of: Informatics

<< Previous Session Proposal | Next Session Proposal >>

IN005 Advancing Artificial Intelligence for Remote Sensing: Overcoming Data Scarcity and Domain Shift

Session Description:

Co-Sponsor(s):

Index Terms:

IN005
Advancing Artificial Intelligence for Remote Sensing: Overcoming Data Scarcity and Domain Shift